which version of vs code is cursor 2.0 based on?
Last night, GitHub Copilot unveiled a stunning array of new featuresâcustom agents, subagentsâand theyâve done it without messing up the UI/UX or stripping away existing functionality, instead continually delivering more meaningful tools to users.
Does Copilot support things like a terminal allowlist? Wonder if its time to give it a try.
Yes, my most commonly used /add-active-tab is gone
Copilot is available for free 50 timesâfeel free to give it a try first!
I do wonder if it will support my workload. I am a very heavy agent user, which is why the removal of allowlist has had such a huge impact on me. Iâve been researching VSCode Copilot, but so far there donât seem to be many people using its agent heavily. So I donât have a lot to go off of⌠Seems like it is mostly autocompletion (whcih I used Github Copilot for in the past.) I am WAY beyond autocompletion now. Which is really quite a bummer, as Iâm starting to feel like I am locked into an IDE that the authors of donât give a â â â â about, and are just marching forward with features without giving their users ANY warning about, NO time to even understand changes, let alone give any kind of reasonable explanation for why changes might be PROPOSED, let alone enacted before users even know what it was that just smacked em upside the head.
Cursor has just had me furious all day today. Its not a pleasant experience anymore, but after checking out Windsurf Cascade, using Claude Code at the terminal which, was better in many ways and not in others, sadly I donât really care for the current state of the agentic IDE market. Claude Code has plugins for JetBrains and VSCode, and I might give that a try. Found something called Kiro AI, which I might give a try, but it looks very primitive.
As a paying Cursor customer, Iâm really quite miffed at the way they are approaching software development for this product. Iâm not some like-chasing socialmedialite. I CARE about the features of the tools I use, and when I become invested in a feature that SAVES ME A CRAPLOAD OF TIME, it is velocity-devastating to lose it on a dime. We have asked for quality changelogs, and we havenât received them. We asked for more stable features, and all we seem to get are more and MORE features, but not stable ones. We have this CONSTANT feature churn and redesign of the IDE, day in and day out, and it is NOT conducive to my PROFESSIONAL WORK.
I really hope someone, takes advantage of the mess Cursor is making right now, and makes something truly better. Now is the opportune time! I fear very soon, this product is just going to be a cesspool of poorly thought out design, usability, and half-built features that are in a constant, never-ending flux, making my day to day job a hellish nitemare. Someone, please, swoop in and make a better, more stable, more reliable product. Please! Because right now, Cursor just aint!
3 Likes
Funny enough, I specifically tried updating to the latest versionâand to fix the issue where moving the Primary Side Bar to the left caused it to revert to the right on next launchâthey simply disabled the âMove Primary Side Bar Leftâ function altogether.
Not a happy camper right now. Not happy at all. Iâm back on 1.7.28 now, because that seems to be the ONLY version where I still have the allowlist, and the agent terminal doesnât hang on every single command. But it still has plenty of other bugs. And every time Cursor crashes, and it crashes a lot, it auto-upgrades me again. Sick and tired of that!
NOT a happy camper. AT. ALL.
1 Like
Hey there, sorry for the confusion here. Since this is on Nightly, we havenât yet had the chance to publish our stable changelog with more details. That will happen likely tomorrow.
To answer your question, we have removed some low-utilized features, but mostly the changes to the context menu and @ symbols still exist â just that you can ask the agent directly for what you need.
The models have both gotten better at tool calling as well as self-gathering context. For example, rather than @git, the models can self-gather git history or use web search tools (instead of needing an explicit @web tag).
In terms of migration plans, for most of these, you can just ask the agent directly for what you need. There are a few that have been deprecated, though. Notably notepads, which we have a separate forum post for here.
@Web should function like a subagent or custom commandârather than relying solely on the modelâs capabilities, since some models tend to struggle in this area.
1 Like
@andrewh Whatâs going on with allowlists in the terminal? The sandbox is unusable for my use case, and having to manually check and allow/deny every single command is simply non-viable for day to day use. Please tell me you guys can add allowlist backâŚ
I donât know why Cursor is removing features, that is frustrating. I did find a sort of workaround, if you want to give it a try.
- type â/â and click âCreate Commandâ
- call the command something like âadd-active-tabâ
- type âadd active tab to contextâ and save
- now you should see the custom command after typing â/â
- Results
Note:
In an empty chat, you can type a request asking what the context is. It usually is the current active tab and the line number your cursor is at, all open files, and recently modified parts of code. So in theory, if you make a request it will assume you mean the current tab, unless its not a fresh chat then it would put weight on previous chats context. But I think one could say âin the active tabâŚâ in plain language and it would add it to the context.
6 Likes
Could you share more? The sandbox should only run more commands and also respect your allowlist
The sandbox is read only, which makes it completely non-viable for my use case. I need a read/write allowlist-based terminal mode for my use case.
1 Like
The sandbox later will try your allowlist as well
Have they already fixed the file corruption when they are thousand lines long?
The sandbox trying the allowlist isnât the problem. The sandbox being READONLY is the problem. My usage doesnât work with a readonly terminal instance. I need a read-WRITE mode that supports an allowlist, which is exactly what I had, until you guys arbitrarily removed the ONLY terminal mode I am willing and able to use.
The sandbox is not a drop-in replacement for the allowlist mode we had in the terminal before you REPLACED that mode with sandbox. Sandbox is a read-only mode, and even if it uses an allowlist (which doesnât really make senseâŚif the sandbox is readonly, then why bother with the allowlist?), it still wonât work for my use case.
This change is deal-breaking for me with Cursor. I cannot use cursor without a READ/WRITE allowlist mode for the agent terminal. The other options just donât fit my use case or otherwise greatly expand the babysitting effort for every agent chat I have running (and, when you have to babysit a chat, you really cannot run more than one at once.)
I have no interest in a Nanny IDE, trying to protect me from myself. I am a power user. I donât want to be coddled and hand-held. I want features that will empower me to ACCELERATE, PRODUCE, and maintain a high performance velocity. Cursor WAS that. Cursor is very rapidly becoming something else, given all the broken features and now the loss of non-sandbox allowlist mode for the agent terminalâŚ
Do you guys understand?
2 Likes
They removed the whole context bar??? This has really broken our entire dev strategy. We have 5 devs on the ultra plan that just had to take huge step back. We can no longer get the results we were getting. Undo Undo!
6 Likes
How do I get the independent agent window like before? It used to be a simple ctrl + shift + p but now I can only get it on the left of the window which is completely breaking my flow, I have a very large monitor and having the primary sidebar on the complete right makes it super annoying.
Edit:
- Allow us to pick WHERE we want the agent window instead of forcing it on us.
- Had to restart 5 times because I couldnât type in the agent window
- You guys say itâs still in early access but that update was forced on me while Iâm on Default update access.
5 Likes
Thatâs great to hear, does this mean the Github integration issues are fixed in 2.0?